THE 1996 WELFARE reform law wrought dramatic change. It reduced the rolls from 4.6 million families to 1.7 million by 2009. Child poverty rates fell and single-mother employment rates went up.

Mitt Romney and his campaign claim the Obama administration has gutted this landmark law. Says a 30-second ad paid for by the Republican National Committee: “Under Obama’s plan, you wouldn’t have to work and you wouldn’t have to train for a job. They just send you your welfare check. Welfare-to-work goes back to being plain old welfare.”

This is false. The disputed July 12 memorandum from the Department of Health and Human Services offers to waive certain provisions in the current law for states that want to try new methods of meeting the work requirement. Even if those waivers did “gut” the work requirement, the main pillars of reform — a fixed federal block grant to the states of $16.5 billion per year, a five-year lifetime limit on federally funded benefits and aggressive caseload reduction goals — would still stand. They would prevent a return to the bad old days of unconditional, lifelong welfare dependency. Of course, given the states’ budget woes, it’s a mystery why any governor would seek to do that in the first place.

Mr. Romney’s sloppy, hyperbolic attack is doubly unfortunate given that welfare was historically a racially fraught issue that reform defused. He and other critics are on firmer ground, however, when they question HHS’s authority to change welfare policy, sweepingly or modestly, through a mere executive-branch memorandum.

There’s a history here. The 1996 reform effectively ended debate on whether to require work; but arguments over exactly what activity counts as work, and how to measure or monitor it, raged on. Competing experts dispute whether it’s more effective to move beneficiaries immediately into a job — the “work first” approach — or to allow more time for education, training or substance-abuse treatment – “human capital development.”

When Congress reauthorized welfare reform in 2005, it took into account data supporting “work first” — as well as a Government Accountability Office report showing that some beneficiaries were getting work credit for dubious activities such as journaling. As a result, the measure, which President Bush signed, included more specific work definitions, coupled with stricter reporting requirements for the states. Additionally, it told states that they would be held accountable for keeping caseloads below the 2005 level, not the 1996 level, as the previous law did.

States chafed at the new norms, arguing that the administrative burdens made it hard to meet the ambitious new caseload reduction norms. There is some evidence that they’re right. But Congress has failed to agree on a new version of the law, so the 2005 version has been repeatedly extended.

Broadly speaking, the HHS memo embodies a policy approach that critics of the 2005 changes advocated before Congress during the most recent reauthorization attempts, but which Congress, for better or worse, did not adopt. It emphasizes experiments with “human capital”-like approaches — reflecting the new administrative reality in the states and recent evidence from pilot projects in Portland, Ore.

It might work; or, as the more responsible critics fear, waivers might dilute the strong pro-work incentives under existing law. We can’t really know until an actual waiver gets granted and a project runs its course. Either way, HHS’s legal justification — that the secretary’s authority to waive state reports on how to meet the work requirement also permits her to let states redefine the work requirement — strikes us as hard to square with Congress’s intent.